Title :
Automatic detection of microvascular obstruction in patients with myocardial infarction
Author :
Trygve Eftest?l;Erlend Singsaas;Kjersti Engan;Leik Woie;Stein ?rn
Author_Institution :
Faculty of Science and Technology, University of Stavanger, Norway
Abstract :
In this study we present a method for segmenting microvascular obstruction in patients with myocardial infarction. The presence of microvascular obstruction is an important prognostic indicator. In late enhanced cardiac magnetic resonance images scar will have very high signal intensity while areas of microvascular obstruction will appear with low signal intensity within the infarction. The method was developed and tested on images from 22 patients. Candidate micriovascular regions within the scar were determined by using adaptive thresholding and training a classifier to distinguish true microvascular obstruction regions from false ones. The best performing classifier (mean(std.dev.)) came out with true positive and negative rates of of 0.91(0.09), and 0.83(0.03) respectively. The results of these preliminary experiments indicate that automatic detection of microvascular obstruction areas is feasible.
Keywords :
"Image segmentation","Myocardium","Injuries"
Conference_Titel :
Computing in Cardiology Conference (CinC), 2015
Print_ISBN :
978-1-5090-0685-4
Electronic_ISBN :
2325-887X
DOI :
10.1109/CIC.2015.7411021